Lithium-ion e-scooter battery engineering: electrochemistry, BMS, thermal runaway, safety standards and life cycle

19.05.2026 Scootify 16 min read

The guide «Charging your battery and caring for it» describes the behavioural and operational side: the 20–80 % window, temperature thresholds of smart chargers, FDNY storage protocol, smart chargers with 80 / 90 % cutoff. This article is an engineering deep-dive into the electrochemistry itself, BMS architecture and thermal runaway physics: why graphite-LiCoO₂ yields a 3.7 V nominal cell, while LFP gives 3.2 V; how Li⁺ intercalation into the anode lattice works molecularly and why the SEI layer on the anode is both your best friend (protects against spontaneous electrolyte decomposition) and your worst enemy (consumes 5–15 % of capacity over service life); why separator melt at 130 °C is the detonator of thermal runaway, not its consequence; how the BMS solves the SoC-estimation problem and why coulomb counting accumulates 5–10 % error per week. This is a separate engineering discipline, paralleling protective gear engineering, braking technique, throttle control — the applied-physics circuit of riding skills is complemented by an engineering circuit for critical subsystems of the scooter.

Prerequisite — understanding of battery architecture and real-world range (Wh, chemistry, cycles), controllers, BMS and electronics (topology, FOC, telemetry), winter operation (BMS charge-block physics below 0 °C) and motors (where that energy goes).

1. Electrochemistry: why 3.7 V and why lithium

Every Li-ion cell is a galvanic cell with two electrodes separated by a porous separator soaked in liquid electrolyte. Discharge is a flow of Li⁺ ions through the electrolyte from the negative electrode (anode) to the positive electrode (cathode), and a simultaneous flow of electrons through the external circuit — that’s the current you draw to the motor.

The physical quantity that defines the cell’s nominal voltage is the electrochemical potential difference of cathode and anode. For the graphite anode (the standard material in 99 % of modern batteries) the equilibrium potential is around 0.1 V vs Li-metal; cathodes have varying potential depending on chemistry:

LiCoO₂ (LCO, Lithium Cobalt Oxide) — 3.9 V vs Li, so cell nominal ≈3.8 V. The first commercial Li-ion (Sony, 1991), high specific energy 150–200 Wh/kg, but thermally unstable (decomposes at ~200 °C, exothermic oxygen release) — used in smartphones and laptops, not in e-scooter battery packs for safety reasons.
LiNi₀.₈Co₀.₁Mn₀.₁O₂ (NMC 811) and LiNi₀.₈Co₀.₁₅Al₀.₀₅O₂ (NCA) — 3.7 V nominal, 200–270 Wh/kg specific energy. The standard of modern premium-segment e-scooter (Apollo Phantom, NAMI, Dualtron Thunder). Cobalt content has been reduced for cost and dependency reasons, nickel increased for energy; thermo-stability around 200 °C (better than LCO, worse than LFP).
LiFePO₄ (LFP, Lithium Iron Phosphate) — 3.2 V nominal (lower because of iron electrochemistry), specific energy only 90–160 Wh/kg, but the olivine structure is extraordinarily thermally stable — runaway threshold ~270 °C and vents without oxygen-exothermic breakdown. Standard of fleet e-scooter, e-mopeds, BYD electric cars and ESS systems; entering e-scooter mainstream (Segway-Ninebot Max G2, Apollo Pro 60 V LFP).
LiMn₂O₄ (LMO, spinel) — 3.7 V, 100–150 Wh/kg, cheaper than NMC, but degrades faster through manganese dissolution into the electrolyte above 50 °C. Found in gen-1 e-scooters and power tools; in modern packs often blended with NMC (Nissan Leaf gen 1).
Li₄Ti₅O₁₂ (LTO, Lithium Titanate) — 2.4 V nominal (anode, not cathode — graphite replacement!), specific energy 60–80 Wh/kg (low), but cycle life >10 000 cycles and operating range −30…+55 °C. A niche chemistry for public transport (Toshiba SCiB, Proterra buses); barely used in e-scooter due to low gravimetric energy.

Base compendium of chemistries — Wikipedia § Lithium-ion battery, Battery University BU-205 «Types of Lithium-ion». The trade-off between specific energy (Wh/kg), specific power (W/kg), cycle life, thermal stability and cost is fundamental — no chemistry wins on every axis.

Intercalation and de-intercalation is the molecular mechanism of charging and discharging. A Li⁺ ion in the charged state sits inside the cathode structure (between layers or in a 3-D framework), and after discharge moves into the anode graphite lattice (between layers). Graphite can accept 1 Li ion per 6 carbon atoms (intercalated formula LiC₆), giving theoretical anode capacity of 372 mAh/g. The cathode (NMC 811) gives a theoretical ~280 mAh/g, so the real cell is cathode-limited, and that’s the fundamental reason all chemistry improvements over the past 20 years concern the cathode, not the anode.

The electrolyte is a solution of a lithium salt (typically LiPF₆) in organic carbonates (EC + DMC + DEC) with small additives that stabilise the SEI on the anode. Non-aqueous — any trace of water reacts with LiPF₆ to form HF and degrades the cell. That’s why pack sealing and separator hermeticity are critical.

The separator is a porous microfilm (PE or PE-PP-PE trilayer) 10–25 µm thick that allows Li⁺ to pass through but electrically isolates anode from cathode. If the separator is punctured (mechanically — a sharp object; chemically — a lithium dendrite grown by fast charging in cold), the cathode and anode touch directly, a localised short circuit appears, the temperature at the contact point spikes to >200 °C — and thermal runaway begins.

2. Cell formats: 18650, 21700, 26650, pouch and prismatic

An e-scooter pack consists of dozens of individual cylindrical or prismatic cells, connected in series-parallel topology. The cell format determines how many cells are needed for the desired Volt-Amp-hours and what the pack’s thermal conductivity will be.

Format	Size	Volume	Typical capacity	Volumetric density	Chemistry	Application
18650	18×65 mm	16.5 mL	2 500–3 600 mAh (NMC), 1 500–1 800 (LFP)	600–700 Wh/L	NMC, LCO, LFP, LMO	Tesla Roadster, laptops, mid-range e-scooter
21700	21×70 mm	24.3 mL	4 000–5 000 mAh (NMC)	700–750 Wh/L	NMC, NCA	Tesla Model 3+/Y, premium e-scooter (Apollo Phantom, NAMI)
26650	26×65 mm	34.5 mL	4 500–5 500 mAh (LFP)	500–600 Wh/L	mostly LFP	Tail-era fleet e-scooter, LFP packs
4680	46×80 mm	133 mL	22 000–26 000 mAh	700–800 Wh/L	NMC, NCA	Tesla Model Y/Cybertruck (2022+), scaling into e-mobility
Pouch	varies (mm)	varies	5–60 Ah	550–650 Wh/L	NMC, LFP	EVs, premium e-bikes, drones
Prismatic	fixed aluminium	varies	50–300 Ah	450–550 Wh/L	LFP, NMC	BYD Blade, EVs, ESS

Cylindrical (18650 / 21700 / 26650 / 4680) is the most widespread format in e-scooter, because they are:

Mass-manufacturable (Panasonic, LG, Samsung, Tesla 4680) — tens of billions of cells per year.
Radially rigid — the metal can withstands internal vent-gas pressure without bulging (unlike pouch).
Standardised — spot-welding and balance-tap pack design is predictable.
Have natural inter-cell space for air or liquid cooling — cylinders in a hex pack leave 10–15 % of volume free for ventilation.

Pouch format (as in Bluetti, EcoFlow, DJI Mavic drones) has +10–15 % volumetric density thanks to the absence of a metal can and maximum geometric packing efficiency of electrodes, but:

Swelling risk — under overheating or degradation a pouch swells, which in an e-scooter pack in a metal deck can destroy the structure.
Harder interconnection — tab welding tab-to-busbar instead of cylindrical spot welds.
Worse heat dissipation — flat geometry is poor at shedding heat without active cooling.

Prismatic format with aluminium can (BYD Blade, CATL) is a compromise between cylindrical and pouch: high structural rigidity, better heat dissipation than pouch, volumetric density between the two. Still rare in the e-scooter segment, but dominates in EVs.

Format overview — Wikipedia § Lithium-ion battery § Format, Battery University BU-301 «A look at old and new battery packaging». The trade-off between cylindrical and pouch is a trade-off between stability + manufacturability and specific volume + design freedom.

3. Pack architecture: series-parallel topology and why 36 / 48 / 52 / 60 / 72 V

An e-scooter battery pack is n cells in series (S) and m cells in parallel (P), where S defines pack voltage and P defines capacity and maximum discharge current. Standard notation is <S>S<P>P:

10S2P — 10 series × 2 parallel = 20 cells. For NMC 3.7 V × 10 = 37 V nominal (4.2 × 10 = 42 V full charge, 2.7 × 10 = 27 V cut-off). This is the standard 36-volt pack of mid-range e-scooters (Xiaomi M365 / Pro / 4 Pro): with 18650 NMC 2 900 mAh × 2P = 5 800 mAh, total 36 × 5.8 = ~210 Wh.
13S3P — 13 × 3 = 39 cells. NMC 3.7 × 13 = 48.1 V (54.6 full / 35.1 cut-off). Apollo City, Niu KQi3 Pro, Segway-Ninebot Max G30: 18650 × 3P = 8 700 mAh, total 48 × 8.7 = ~420 Wh.
14S4P — 14 × 4 = 56 cells. NMC 3.7 × 14 = 51.8 V (“52 V”). Apollo Air Pro, Dualtron Mini: 21700 × 4P = 16 000 mAh, total 52 × 16 = ~830 Wh.
16S5P — 16 × 5 = 80 cells. NMC 3.7 × 16 = 59.2 V (“60 V”). Apollo Phantom V3 60 V, NAMI Burn-E: 21700 × 5P = 25 000 mAh, total 60 × 25 = 1 500 Wh.
20S6P — 20 × 6 = 120 cells. NMC 3.7 × 20 = 74 V (“72 V”). Dualtron Thunder 3, Wolf King GTR: 21700 × 6P = 30 000 mAh, total 72 × 30 = 2 160 Wh.

Standard voltages 36 / 48 / 52 / 60 / 72 V aren’t an accident — they’re historical heritage from the lead-acid era (multiples of 12 V) adopted as the reference workflow for controllers and motor windings. Each voltage step allows you to lower the current for the same power (P = U × I): a 72-volt 30-amp pack delivers 2.16 kW at the same 30 A, while a 36-volt one delivers only 1.08 kW. Lower current means thinner conductors, less I²R loss in the motor, higher theoretical efficiency. That’s why the performance segment trends towards 60–72 V.

Why series-parallel and not just series? One 18650 NMC 3 Ah cell can deliver 20–30 A maximum on short discharges (typical 30 A for INR18650-30Q). If 60 A is needed (for a dual-motor 60 V × 60 A = 3.6 kW), one S stack gives only 20–30 A — so P=2 parallel cells are needed. P also multiplies pack capacity proportionally (3 Ah × 2P = 6 Ah) and extends life, because each cell takes a proportionally smaller share of the cycle.

The P trade-off — adds mass and cost linearly. So the pack designer solves the puzzle (needed capacity) ∩ (needed continuous discharge current) ∩ (mass budget) ∩ (thermal budget) — and picks the minimally sufficient P.

4. The SEI layer: your best friend and your worst enemy

Solid Electrolyte Interphase (SEI) is a thin (5–50 nm) layer of lithium-containing salts (Li₂CO₃, LiF, ROCO₂Li, ROLi) that forms on the graphite anode surface during the first cycles of a new battery. It builds through electrochemical reduction of the electrolyte (carbonates) on the anode surface at potentials below ~1.0 V vs Li⁺/Li (i.e. at every graphite charge).

This layer simultaneously solves two problems and creates one:

Problem 1: prevents direct contact between anode and electrolyte. Without SEI the electrolyte would continuously react with charged graphite (graphite potential below the thermodynamic stability range of carbonates), and the battery would self-discharge with gas evolution and degradation. SEI is a necessary bandage.

Problem 2: passes Li⁺ through itself. SEI is electrically insulating but ionically conductive; a Li ion passes through it at a rate that defines the cell’s rate capability. A good SEI is thin and uniform; a degraded one is thick and patchy.

The problem: SEI grows with every cycle. Each charge «eats» 0.1–0.5 % of capacity for forming new SEI that covers freshly exposed graphite through structural micro-changes. Over 500 cycles this accumulates 5–15 % capacity loss — and this is one of the two main mechanisms of capacity fade (the second is cathode degradation). SEI also grows in thickness over time (calendar aging), increasing the cell’s internal resistance — the second end-of-life criterion.

What accelerates SEI growth:

High temperature — Arrhenius kinetics: every doubling of temperature above 25 °C roughly doubles the growth rate. Storing the pack at 40 °C gives 2× degradation vs 25 °C.
High anode voltage (meaning high pack SoC) — full charge keeps graphite at the lowest potential where electrolyte is maximally unstable. This is the physical reason behind the rule “don’t keep a pack at 100 % SoC for long” — SEI grows faster.
Fast charging (high C-rate) — uneven Li deposition on graphite creates gradient stress, forms dendrites and new SEI sites.
Charging below 0 °C — Li intercalation into graphite slows down faster than metal plating on the surface, so at <0 °C metallic Li dendrites form that pierce the separator → catastrophic failure.

Deep compendium of SEI mechanisms — Ossila — Introduction to the Solid Electrolyte Interphase (SEI) Layer, Battery University BU-808 «How to prolong lithium-based batteries», Edström et al. «The cathode-electrolyte interface in the Li-ion battery» review in Electrochimica Acta (2004). SEI is why literally no Li-ion battery «keeps its full capacity» — it degrades continuously from the moment of manufacture, even sitting on a shelf.

5. BMS architecture: protection, balancing and SoC-estimation

Battery Management System (BMS) is an electronic board (often as thick as a credit card) that monitors every series-level group of the pack and controls two pairs of MOSFET switches (charge + discharge) for cutting off the pack in emergency conditions. In an e-scooter pack the BMS lives inside the pack itself, wired to each P-group via balance-tap leads.

BMS protection functions:

Over-voltage cutoff (OVP) — limits maximum series-group voltage to ~4.20–4.25 V for NMC, 3.60–3.65 V for LFP. If one group exceeds — the BMS shuts the charge MOSFET, charging halts.
Under-voltage cutoff (UVP) — limits minimum voltage to 2.5–2.8 V for NMC (3.0–3.1 V for LFP). If one group drops below — the BMS shuts the discharge MOSFET, the motor stops. UVP is critical: deep discharge below 2.0 V destroys the SEI and leads to internal short circuit on the next charge (and that’s the typical death of a pack that lay “discharged” in the garage for a year).
Over-current cutoff (OCP) — limits maximum continuous + peak current. On the 36-volt Xiaomi pack OCP is typically 30 A continuous / 50 A peak; on the 72-volt Dualtron — 80–120 A.
Over-temperature cutoff (OTP) — NTC thermistor near the hottest pack zones. Trips at 60–70 °C for discharge / 45–55 °C for charge (lower because the exothermic intercalation reaction adds 5–10 °C).
Short-circuit cutoff (SCP) — trips in microseconds when I > 200 A.

Cell balancing — passive vs active.

No two cylindrical cells off the factory are identical: capacity scatter ±2–3 %, internal resistance scatter ±5–10 %. In a pack of 80 cells (16S5P) this means after 50–100 cycles one series group runs ahead (higher SoC after charge) — and it’s that group that will hit OVP cutoff first, stopping pack charging before the others reach 100 %. Effective pack capacity = capacity of the weakest group, not the average.

Passive balancing — simple and cheap: when the BMS sees that one group is above the average by >50 mV, it opens a parallel resistor across that group (typically 50–100 Ω, sinking 30–60 mA), dissipating energy as heat. Triggers only at the end of charging, in the CV phase, when there’s time to “wait out” the equalisation. Balance current 50 mA × 2 h is only 100 mAh of equalisation, so with a pack capacity of 8 000 mAh dozens of cycles are needed to compensate 2 % scatter. Almost all e-scooter BMS are passive, because it’s cheap and covers the real use case.
Active balancing — more complex: the BMS shuttles charge from higher groups to lower groups via inductor / transformer / capacitor pump (efficiency 70–90 %). Balance current 200–500 mA, so equalisation is significantly faster and there’s no thermal loss. Used in EVs, ESS, expensive e-bike packs; in e-scooter — rarely, due to a cost penalty of +$20–40 per pack.

State of Charge (SoC) estimation — three methods:

Coulomb counting — integration of current through the pack over time: SoC(t) = SoC(0) + ∫I dt / Capacity. Accurate short-term but accumulates error due to shunt-resistor precision and temperature-coefficient drift (±0.5 % per day, ±5–10 % per week). Needs regular recalibration on full charge or full discharge.
Open Circuit Voltage (OCV) lookup — after 30+ min of rest the pack voltage settles at an equilibrium that maps unambiguously to SoC via the chemistry’s OCV curve (NMC: 4.20 V = 100 %, 3.90 V = 80 %, 3.70 V = 50 %, 3.30 V = 20 %, 2.80 V = 0 %). Very accurate (±1 %), but works only at rest.
Kalman filter / extended Kalman filter (EKF) — hybrid: combines coulomb counting (short-term precision) with OCV (long-term drift correction) + an electro-thermal model of internal resistance. Standard in EVs, starting to appear in premium e-scooter BMSes. Continuous ±1–2 % accuracy without needing full discharge for recalibration.

State of Health (SoH) — capacity as a percentage of nominal: SoH = current measured capacity / rated capacity. End-of-life pack — typically at SoH = 80 % (industry standard; in some Tesla marketing campaigns — 70 %). SoH is measured via a full discharge cycle, which is hard for users — so BMS apps often do an estimate based on internal resistance growth (R growth correlates with SoH degradation, but noisily).

Detailed BMS-function overview — Texas Instruments «Battery management systems» application note (slvae08), Wikipedia § Battery management system, Plett «Battery Management Systems, Volume I: Battery Modeling» review (Artech House, 2015).

6. Thermal runaway: exothermic cascade physics and propagation prevention

Thermal runaway is a self-amplifying exothermic chain reaction inside a Li-ion cell that occurs when the internal temperature exceeds a critical threshold (~80 °C for NMC; ~270 °C for LFP), beyond which the rate of heat generation from chemical reactions exceeds the rate of heat dissipation from the cell, the temperature keeps rising, and the cell enters catastrophic failure mode with vented hot gases, flames, in extreme cases detonation.

Thermal runaway stages (for NMC, from Feng et al. «Thermal runaway mechanism of lithium ion battery for electric vehicles: A review» in Energy Storage Materials 2018):

Temperature	What happens	Heat-release rate
25–80 °C	Normal operation. SEI is stable.	0 (thermal equilibrium)
80–120 °C	SEI decomposition. The SEI layer begins to break down, exposing graphite. The anode reacts with the electrolyte, exothermic ~250 J/g.	~0.1 W/g
120–150 °C	Separator melt. PE separator melts at 130 °C (shut-down feature: pores close, ion flow stops). If the separator is thick enough — the pack survives. If shut-down alone isn’t enough — the cell keeps heating.	~0.5 W/g
150–200 °C	PVDF binder decomposition + cathode-electrolyte reaction. The cathode oxide begins releasing oxygen into the electrolyte, which oxidises exothermically.	~5 W/g
200–250 °C	Cathode breakdown. Oxygen is massively released from the cathode (NMC), exothermic > 1 000 J/g. Electrolyte burns. The vent valve fires, hot gases (H₂, CO, CO₂, CH₄, HF) escape.	>50 W/g
250–800 °C	Full thermal runaway. Vent flame, possible detonation. Pack-neighbouring cells receive radiative + conductive heat and start their own runaway — propagation.	>500 W/g

Thermal runaway triggers:

Internal short circuit. Lithium dendrite, micro-perforation of the separator from a manufacturing defect, mechanical crush (car impact), nail penetration (UL 2271 / UN 38.3 test). Local short → point temperature >200 °C → cascade.
External short circuit. Pack output shorted directly — current >1 000 A within microseconds, I²R heating to >100 °C within seconds. The BMS must trip in microseconds (SCP).
Over-charge. Cell voltage > 4.5 V (for NMC) — lithium plating on anode, excess energy in cathode. Standard UL 2271 test: over-charge to 200 % SoC. A quality BMS stops at 4.25 V.
Over-temperature. Charging at +50 °C + sun-baked car → pack reaches 80 °C → SEI starts to decompose.
Mechanical abuse. Dropping the pack from height, crushing under car wheels, nail penetration (UL nail penetration test).

Propagation prevention — how to stop the cascade after one failed cell:

Cell spacing. A 1–3 mm gap between cylinders slows conductive heat transfer; 10+ mm almost stops propagation.
Heat-absorbing foam / phase-change materials. PCM (e.g. paraffin with high heat-of-fusion) between cells absorbs 200+ J/g at melting, buying tens of seconds for venting.
Ceramic separator. Aluminium oxide coating on PE separator raises the melt temperature to 200+ °C — the cell survives more thermal abuse without internal short.
Vent valve in every cell. An 18650 cylinder has CID (Current Interrupt Device) and a vent valve in the positive terminal — hot gases exit controllably, without detonation.
Pressure-relief and fire-resistant pack housing. Aluminium deck + mica isolation + vent port in the underside. Tested in UL 2271 nail penetration and UL 2272 + UL 2849 system level.
BMS thermal cutoff. On detection of >70 °C the BMS shuts both charge and discharge MOSFETs, isolating the pack from any load. Doesn’t prevent an internally triggered runaway, but disconnects external sources of energy injection.

Deep compendium — Feng et al. «Thermal runaway mechanism of lithium ion battery for electric vehicles: A review» Energy Storage Materials (2018, PMC), Wikipedia § Thermal runaway, Battery University BU-304a «Safety Concerns with Li-ion», Wang et al. «Thermal runaway caused fire and explosion of lithium ion battery» Journal of Power Sources (2012). LFP chemistry has a runaway threshold at 270 °C (vs 80 °C for NMC) and without oxygen release — a fundamental safety advantage that’s making LFP the preferred choice for urban e-scooter where public exposure to thermal incident grows.

7. Full safety standard matrix: UL, EN, IEC, UN

Safety certification of a lithium-ion e-scooter battery is five independent layers covering the cell, the pack, the system, type approval and transport. In NYC and London, UL 2271 + UL 2272 (or UL 2849 for e-bikes) is mandatory; in Europe — EN 50604-1 (LEV) and EN 17128 (PLEV/e-scooter); FDNY 2024 recorded a 67 % drop in deaths after NYC Local Law 39 introduced mandatory UL standards (NYC.gov FDNY release 2025).

Standard	Level	Region	What it tests	Key tests
IEC 62133-2	Cell-level	World	Safety of an individual cell under normal + reasonably foreseeable misuse	external short, abnormal charge, forced discharge, crush, impact, vibration, thermal abuse, low pressure
UL 2271	Pack-level	USA/Canada	Battery pack for light EV (e-scooter, e-bike, e-skateboard). Pack including BMS.	over-charge, short circuit, drop, crush, nail penetration, vibration, immersion in water, thermal cycling, projectile
UL 2272	System-level	USA/Canada	Personal e-Mobility device (e-scooter, hoverboard). Pack + charger + vehicle integration.	UL 2271 + system-level abnormal use, electrical fault, fire propagation, water exposure
UL 2849	System-level	USA/Canada	E-bike electrical system (motor + battery + controller + charger).	UL 2272 + motor + drivetrain integration; required by NY State Local Law 39 for e-bikes
EN 50604-1	Pack-level	EU	Light EV traction battery pack. European analogue of UL 2271, adapts IEC 62133.	thermal, electrical, mechanical, environmental
EN 17128	System-level	EU	Personal Light Electric Vehicles (PLEV) — e-scooter, hoverboard.	Analogue of UL 2272 for Europe; CE-certification cycle for PLEV
UN 38.3	Transport	World (UNECE)	Transport safety of Li-ion. Required for air, sea and ground freight transport.	T1: altitude simulation, T2: thermal cycling, T3: vibration, T4: shock, T5: external short, T6: impact/crush, T7: overcharge, T8: forced discharge
UN R136	Type approval	UNECE	Type approval for L-category vehicles (mopeds, motorcycles, type-approved e-scooter).	Includes battery safety requirements for homologation

Why NYC LL39 worked: before 2023 NYC allowed e-scooter sales without UL certification; the market was flooded with cheap ignitable packs without BMS protection, without UL nail penetration. FDNY 2023: 19 deaths from Li-ion battery fires, mainly e-bike + e-scooter. NYC Local Law 39 (in force March 2024) mandated UL 2271/2272/2849 for all sales + bans “refurbished” packs without recertification. In 2024 — 6 deaths (−68 %), 277 fires (vs 268 in 2023, −10 % at significantly larger fleet). EN 17128 in Europe was adopted in 2020 and is gradually becoming mandatory in national CE regimes (France from 2024).

UN 38.3 — the transport foundation. Every Li-ion battery shipped by plane, ship or truck across international borders is required to pass 8 UN 38.3 tests (test summary + DGR document Class 9). Airlines verify the UN 38.3 cert before accepting an e-scooter or spare pack in luggage; without it — refusal. That’s why a DIY pack or a repaired pack cannot legally be flown — without UN 38.3 cert.

Standards compendium — UL Solutions «UL 2272 Personal e-Mobility Evaluation, Testing and Certification», UN 38.3 «Recommendations on the Transport of Dangerous Goods, Manual of Tests and Criteria», CEN EN 17128 product page. All seven standards form layered defense: cell-level (IEC 62133) → pack-level (UL 2271 / EN 50604-1) → system-level (UL 2272 / UL 2849 / EN 17128) → transport (UN 38.3) → type approval (UN R136).

8. Life-cycle physics: cycle aging, calendar aging and Arrhenius

Li-ion battery degradation is the sum of two independent mechanisms that superimpose:

Cycle aging — capacity fade as a function of the number of charge-discharge cycles. The principal driver is SEI growth (see Section 4) + loss of active cathode material through intercalation strain. Empirically described by a power-law: Capacity_loss ∝ Cycles^n, where n ≈ 0.5–0.7 depending on chemistry. Typical lifetime figures:

NMC 21700 at 25 °C operating, 1 C charge, 100 % DoD: 500–800 full cycles to 80 % SoH.
NMC 21700 at 25 °C, 1 C charge, 80 % DoD (window 10–90 %): 1 500–2 500 cycles — and that’s the “2× life” effect of the 20–80 rule.
NMC 21700 at 25 °C, 1 C charge, 40 % DoD (window 30–70 %): 3 000–5 000 cycles.
LFP 26650 at 25 °C, 1 C, 100 % DoD: 2 000–4 000 cycles to 80 % SoH (fundamental chemistry advantage).
LFP at 80 % DoD: 6 000–10 000 cycles.
LTO (Toshiba SCiB): >10 000 cycles at 100 % DoD — record-holder.

Depth of Discharge (DoD) effect — non-linear. Going from 100 % → 80 % DoD gives 3× life (not 1.25×), because SEI growth accelerates exponentially with depth of discharge through mechanical strain on graphite. So all smart chargers with 80 / 90 % cutoff and all OEM “don’t go to 100 %” recommendations are not marketing, but physics.

Calendar aging — capacity fade as a function of time + temperature + SoC, independent of cycling. SEI grows at any temperature above 0 °C, but the rate obeys the Arrhenius equation:

k(T) = A × exp(−Eₐ / (R × T))

where Eₐ is the activation energy of SEI growth (~50–80 kJ/mol for NMC), R is the universal gas constant, T is absolute temperature (K). Practically — a doubling of degradation rate per +10 °C (10–15 °C depending on chemistry).

Plotted against SoC, calendar aging looks like this (NMC 21700, 1 year of storage, Battery University BU-702 «How to store batteries»):

Storage SoC	Temperature	Capacity loss per year
100 %	0 °C	6 %
100 %	25 °C	20 %
100 %	40 °C	35 %
40 %	0 °C	2 %
40 %	25 °C	4 %
40 %	40 °C	15 %

So the storage protocol is: 40–60 % SoC + room temperature (15–25 °C), with top-up 1–2 times per month to avoid deep discharge. More — in the charging and care guide.

Internal resistance growth — the second end-of-life criterion. Alongside capacity fade, a deluxe pack degrades through DC internal resistance growth by 50–100 % over 500–1 000 cycles (via SEI thickening + cathode-electrolyte interface degradation + binder breakdown). This means growing I²R losses at the same discharge and increased heat generation under load. A pack with SoH 80 % and R_int 2× normal may have normal at-rest voltage, but trip on UVP during a hill climb through accelerated voltage sag — and that’s the typical “battery is end-of-life, but voltage normal” symptom.

Compendium of lifetime mechanisms — Battery University BU-808 «How to prolong lithium-based batteries», Vetter et al. «Ageing mechanisms in lithium-ion batteries» Journal of Power Sources (2005), Pinson & Bazant «Theory of SEI formation in rechargeable batteries» Journal of the Electrochemical Society (2013).

9. What this deep-dive means for daily practice

Engineering isn’t purely academic. Concrete take-aways for an e-scooter owner:

LFP chemistry, where available, is objectively safer. When choosing between two models at the same price point — one NMC, the other LFP — LFP wins on runaway threshold (270 vs 80 °C) and cycle life (3 000 vs 800), loses on specific energy (~30 % heavier pack for the same Wh). For daily commuting in a European or North American city — LFP is almost always traceable.
A 60 / 72 V pack with 21700 cells wins in the performance segment not only on raw power, but also through lower continuous current per cell → less I²R heating → lower SEI stress → longer pack life.
UL 2271/2272/2849 certification is not marketing. A non-certified pack cannot legally be sold in NYC, and that’s not accidental — FDNY 2024 statistics show −68 % deaths. If an OEM can’t show a UL test report — that’s a signal.
Charge protocol matters more than storage temperature. 80 % cutoff on the daily charge plus 40–60 % SoC in seasonal storage together give 3–5× pack life vs the naive “charged to 100 %, parked in the garage until spring”.
BMS is the pack’s heart. When DIY-building or repairing a pack you can’t substitute the OEM BMS with a “generic 30A 13S” from AliExpress — that often means no balance taps, primitive coulomb counting, no thermal cutoff. The pack will work, but degrade within 100 cycles and carry increased runaway risk.
Charging in the cold is pure damage. Below 0 °C Li dendrites grow deterministically; a BMS without a low-temperature lock-out (typical of cheap packs) will let you do it. The manual for a quality pack — pre-warm the pack in a warm room to 10+ °C before charging.
End-of-life is not “battery stopped working” — it’s 80 % SoH or 2× internal resistance. Symptom — distance covered is 25 % of nominal range, voltage sag under load, sudden UVP cuts on climbs. At this point the pack still works, but is no longer safe for aggressive charging (faster SEI degradation on a thin remaining cycle envelope).

10. 8-point recap for an engineering mindset

NMC 200–270 Wh/kg vs LFP 90–160 Wh/kg — a specific-energy vs cycle-life + safety trade-off. The sweet-spot chemistry isn’t a stationary choice but an engineering decision under a concrete use case.
18650 / 21700 / 26650 / pouch / prismatic — formats with different trade-offs between density, manufacturability and stability. Cylinders with vent valves dominate e-scooter; pouch carries swelling risk; LFP prismatic is going mainstream.
Series-parallel topology n S × m P defines (n) voltage and (m) capacity + maximum continuous current. 13S3P = 48 V mid-range, 20S6P = 72 V performance. Each V step lowers I for the same P, reducing I²R losses.
SEI layer simultaneously protects electrolyte from anode and consumes 5–15 % of capacity over 500 cycles through Arrhenius kinetics; the 20–80 SoC rule minimises SEI growth rate.
BMS = protection (OVP/UVP/OCP/OTP/SCP) + balancing (mostly passive in e-scooter) + SoC-estimation (coulomb counting + OCV + Kalman). SoH 80 % is the industry end-of-life criterion.
Thermal runaway — exothermic cascade at ~80 °C (NMC) / 270 °C (LFP). Triggers: internal short (dendrite, crush), over-charge, over-temperature. Propagation prevention: cell spacing, ceramic separator, vent valves, pack housing.
5-layer safety certification: IEC 62133 (cell) → UL 2271 / EN 50604-1 (pack) → UL 2272 / UL 2849 / EN 17128 (system) → UN 38.3 (transport) → UN R136 (type approval). NYC Local Law 39 from 2024 shows −68 % deaths after mandatory UL certification.
Life is the integral of cycle aging (DoD-dependent) + calendar aging (Arrhenius, exponential in temperature and SoC). Storage protocol 40–60 % SoC + room temperature gives 5+× longer calendar life vs 100 % SoC at 40 °C.

Lithium-ion battery engineering is five intersecting disciplines (electrochemistry, materials science, electronics, thermal physics, regulatory engineering), each with its own independent trade-off space. This material is not meant to make you open the pack and start refluxing electrolyte, but for behavioural rules from the charging guide to acquire physical meaning: when a smart charger limits at 80 %, it isn’t “saving you money” — it’s physically slowing SEI growth. When the BMS blocks charging at −2 °C, it physically prevents formation of lithium dendrites that would pierce the separator on the next cycle and trigger thermal runaway. When the OEM shows a UL 2271/2272 cert, it has physically passed nail-penetration and over-charge tests where pack-neighbouring cells are tested for propagation.